关键词: 府际合作, 科技创新政策, 多层网络, 政策网络

Abstract:     The diversification of innovation subjects and the increasing span of innovation fields make the governance of collaborative innovation more difficult. How to promote innovation policy coordination through intergovernmental cooperation and provide institutional guarantee for collaborative innovation becomes particularly important. As the premise of optimizing resource allocation and realizing intergovernmental coordination, intergovernmental cooperation can provide an important guarantee for the effective and orderly operation of public affairs management and the realization of public interests. However, inter-governmental cooperation not only has the impetus but also faces heavy resistance in the process of policy formulation and implementation due to the complex environment of public governance, the pursuit of the interests of government departments and the institutional constraints outside government departments. Current scholars have analyzed the impetus and resistance of intergovernmental cooperation from the perspective of governance environment and the relationship between policy subjects. The research methods mainly focus on qualitative analysis such as theoretical deduction and grounded theory, but lack quantitative research. The research content explores the power and resistance of inter-governmental cooperation from the public governance environment and the relationship among policy subjects at the same level, and there is a lack of cross-level research. However, the policy network is a complex relationship network formed by multiple subjects in the process of policy formulation and implementation. It is not only dynamic but also hierarchical. The policy network includes horizontal cooperation among central policy subjects or local policy subjects, as well as vertical central-local relations. Different levels of networks are interdependent and influence mutually. Therefore, this paper explores the mechanism of inter-governmental cooperation from a cross-level perspective. 
    This research constructs a multi-level network model of intergovernmental cooperation by using the cooperation and citation relationship among the policy subjects presented in the policy texts promulgated by the Central Government and Beijing from 2012 to 2017. The method of Multilevel Exponential Random Graph Models is used in the model construction. The purpose of the research is to break through the limitation of the traditional single-level intergovernmental cooperation in theory. Placing the central and local governments in the same framework is conducive to enrich the quantitative study of inter-governmental cooperation. It is beneficial to the optimization of cooperation relations at the same level and the governance of cross-level policy response relations, thus improving the comprehensiveness and effectiveness of policy formulation and promoting the smooth implementation of policies in terms of application. The research is mainly based on four dimensions: the characteristics of central policy subject and the cooperation of central policy subjects, the characteristics of policy diffusion of central policy subject and the cooperation of central policy subjects, the characteristics of policy response of local policy subject and the cooperation of local policy subjects, as well as the relational characteristics of central policy subjects and the cooperation of local policy subjects. 
    The study points out that the number of partners, the history of cooperation and the common partners play an important role in promoting cooperation among central policy subjects (Hypothesis H1, H2, H3). In terms of the characteristics of the policy diffusion of the central policy subjects, the smaller administrative influence promotes the cooperation of the central policy subjects (Hypothesis H4). However, the common target of policy diffusion will hinder the cooperation of central policy subjects (Hypothesis H5). In terms of the characteristics of policy response of local policy subjects, policy response and different policy respondents hinder the cooperation of local policy subjects (Hypothesis H6, H7b). However, the common policy respondents will promote the cooperation of local policy subjects (Hypothesis H7a). In terms of the relational characteristics of central policy subjects, the cooperation of central policy subjects promotes the cooperation of local policy subjects (Hypothesis H8), while the number of partners of the central policy subjects hinders the cooperation of the local policy subjects (Hypothesis H9). 
    Subsequently, this paper uses the method of Multilevel Exponential Random Graph Models to construct a multi-level network model of inter-governmental cooperation based on the central and local levels and carries out the empirical analysis. The empirical results show that the proposed multi-level network model of inter-governmental cooperation can reflect the real observation network. They also show that Hypothesis H7b is rejected, Hypothesis H8 is not significant, and other hypotheses are tested. It follows that the policy subjects with more partners and cooperation history are more likely to attract other policy subjects to cooperate with them for the central policy subjects. Policy subjects with great influence on policy diffusion have a promoting effect on the cooperation of policy subjects at the same level. It is easier to form cooperative relations between central policy subjects with common partners and different policy diffusion objects. As far as local policy subjects are concerned, policy similarity promotes cooperation among local policy subjects at the same level. When the response of local policy subjects to the central policy can meet the needs of heterogeneous resource accumulation to overcome their own limitations, cooperation with policy subjects at the same level will be reduced. The number of partners of central policy subjects will inhibit the cooperation of local policy subjects at the same level. However, the promotion effect of cooperation among central policy subjects on cooperation among local policy subjects is not significant. 
     According to the research results, combined with policy formulation and implementation practice, the study puts forward the following countermeasures and suggestions. First, the cooperation among central policy subjects should be strengthened, and the coordinating role of the policy subjects with strong policy effectiveness and high policy similarity in the cooperation among the central policy subjects should be paid attention to in policy formulation. Secondly, the central policy subjects should clarify the characteristics of its policy diffusion, find its own shortcomings and improve itself to increase the scientific and effectiveness of policy formulation in the process of policy implementation. Finally, local policy subjects should strengthen cooperation with policymakers at the same level and respond to the central policy subjects at the same time. 
    The multi-level network analysis model of intergovernmental relations based on Multilevel Exponential Random Graph Models can be further extended to the policy network of prefecture-level cities, county-level cities and other administrative levels to explore more inter-governmental cooperation mechanisms. This can provide a powerful reference for optimizing and governing inter-governmental relations.

Key words: inter-governmental cooperation, science and technology innovation policy, multi-level network, policy network